CHAPTER 3

MATHEMATICAL MODELLING
3.1. MALTHUS MODEL OF
POPULATION
The total population of a country is clearly a
function of time, N(t) [NOTE: N may be measured in millions, so values of N less than 1 are
meaningful!]. Given the population now, can
we predict what it will be in the future?
Suppose that B is a function giving the per
capita birth-rate in a given society, ie B is
the number of babies born per second, divided
1

by the total population of the country at that

moment. Note that B could be small in a big
country and large in a small country - it depends
on whether there is a strong social pressure on
people to get married and have kids. Now B
could depend on time and it could depend on N .
But suppose you dont believe these
things: suppose you think that people will
always have as many kids as they can, no matter
what. Then B is constant. Now just as
Distance = Speed Time
when speed is constant, so also we have
#babies born in time t = BN t
2

Similarly let D be the death rate per capita;

again, it could be a function of t (better medicine,
fewer smokers) or N (overcrowding leads to famine/disease)
but if we assume that it is constant, then
#deaths in time t = DN t
So the change in N , N , during t is
N = #birth #deaths
Provided there is no emigration or immigration. Thus,
N = (B D)N t
and so

N
t

= (B D)N or in the limit as t

0,
dN
= (B D)N = kN
dt
3

(1)

if k = B D.
This model of society was put forward by
Thomas Malthus in 1798. Clearly Malthus
was assuming a socially static society in which
human reproductive behaviour never changes
with time or overcrowding, poverty etc... What
does Malthus model predict? Suppose that the
population now is N , and let t = 0 now.
R dN
R
dN
kdt =
From dt = kN we have
N =
R
k dt = kt + c
so ln(N ) = kt + c and thus N (t) = Aekt.
Since N = N (0) = A, we get:
N (t) = N ekt
with graphs as shown on figure 1.
4

(2)

Figure 1: Graphs of N (t), for different values of k

The population collapses if k < 0 (more deaths

than births per capita), remains stable if (and
only if) k = 0, and it explodes if k > 0
(more births than deaths). Malthus observed
that the population of Europe was increasing, so
he predicted a catastrophic population ex5

plosion; since the food supply could not be

expanded so fast, this would be disastrous.
In fact, this didnt happen (in Europe). So
Malthus model is wrong: many millions went
to the US, many millions died in wars.
Second, the static society assumption has
turned out to be wrong in many societies, with
B and D both declining as time passed after
WW2.

Summary: The Malthus model of population

is based on the idea that per capita birth and
death rated are independent of time and N . It
leads to exponential growth or decay of N .
6

3.2. IMPROVING ON MALTHUS

Malthus model is interesting because it shows
that static behaviour patterns can lead to disaster. But precisely because ekt grows so quickly,
Malthus assumptions must eventually go wrong
- obviously there is a limit to the possible population. Eventually, if we dont control B, then
D will have to increase. So we have to assume
that D is a function of N .
Clearly, D must be an increasing function of
N ... but which function? Well, surely the
simplest possible choice (Remember: al7

ways go for the simple model before trying a

complicated one!) is
(logistic)
D = sN, ASSUMPTION
s = constant

(3)

This represents the idea that, in a world with

finite resources, large N will eventually
cause starvation and disease and so increase D.
Remark: In modelling, it is often useful to
take note of units . Units of D are (#dead
people) / second / (total # people) = (sec)1.
Units of N are # (ie no units). So if D = sN ,
units of s must be (sec)1.
As before, let N be the value of N at t = 0.

We have to solve
dN
= BN DN = BN sN 2
dt
with the condition N (0) = N
We can and will solve this, but lets try to
guess what the solution will look like (a useful
skill - in many other cases you wont be able to
solve exactly!). Suppose that N is very small.
Then (by continuity) N (t) will be very small for
t near to zero.
Of course if N is small, N 2 is much smaller
and can be neglected. [Remember that N may
be measured in millions or billions, so N can be
small.] So at early times, our ODE is almost
9

linear and so
dN
BN N (t) N eBt
dt
So at first the population explodes, as Malthus
predicted. On the other hand, if N continues
to grow, since N 2 grows faster than N , we will
reach a point where sN 2 BN ie N B/S.
2
At that point , since dN
=
BN
SN
, the popdt

ulation will stop growing. So B/S should measure the maximum population possible. So we
guess that the solution should look like this:
ie it starts out exponentially and ends up
approaching B/S asymptotically. The dotted
part is a reasonable guess!
OK, now that we know what to expect, lets
10

actually solve it!

dN
= BN sN 2 t =
dt
Write

1
N (BsN )

dN
+c
N (B sN )

+ BsN

1 = (B sN ) + N
= B + ( s)N 1 = B , = s
= 1/B , = s/B, so

11

dN
dN
dN
1
s
=
+
N (B sN )
B
N
B B sN
1
1
=
ln N ln |B sN |
B
B
Now here we begin to feel uneasy - what if

N = B/s at some time? (ln(0) is not defined).

In fact we should have worried about this when
we first wrote

1
BsN

- how do we know that

we are not dividing by zero?? Lets not worry

about that just now: lets assume (temporarily) that B sN is never zero. That is,
we assume either that N is always either less
than B/s or more than B/s. OK, lets
take less than first. So |B sN | = B sN ,

12

and we get
1
1
t =
ln N ln(B sN ) + c
B
B
N
1
ln
+c
=
B B sN

So
N
BsN

= KeBt. Since N = N (0),

BsN

=K

so
N
N
=
eBt
B sN B sN
Solve for N ,
B

N (t) =
s+

(4)

s eBt

Remark: It is a very good habit to check

that your solution agrees with your assumptions
- to guard against mistakes!
13

Check: N (0) =

B 
s+ B s

= N correct! Check:

If B sN > 0 is true at t = 0 then B sN > 0



B
B
 B
so N
<
s > 0 so
B
s for all t ie
Bt
s+ s e
N

N (t) < B/s which is consistent.

The graph of (4) is easy to sketch:

This is the famous logistic curve; N (t)

given by (4) is called the logistic function;
and

dN
dt

= BN sN 2 is the logistic equa-

tion.
14

Its easy to see what is happening here. Initially the population is small, plenty of food and
space, so we get a Malthusian population explosion. But eventually the death rate rises until it
is almost equal to the birth rate (ie sN B or
N B/s) and then the population approaches
a fixed limit.
This situation is what people usually mean
when they use the word logistic. But we
are not done yet: on page 12 we assumed that
N (t) < B/s. What if N (t) > B/s?

15

Then |B sN | = (B sN ) so:
1
1
ln N ln(sN B) + c
B
B
1
N
=
ln
+c
B sN B

t =

N (t) =

B


B
Bt
s sN
e

(5)

Check N (0) = N and N (t) > B/s

And now the graph is

Again, the meaning is clear: the initial popu16

lation was so big that the death rate exceeded

the birth rate, so of course the population declines until it gets near to the long-term sustainable value.
The number B/s is called the carrying
capacity or the sustainable population
- in all cases, it is the value approached by N (t)
as t . If we set
N = B/s

(6)

then our solutions are:

N

N (t) =
1+
N (t) =

(N < N) (7)

(N > N) (8)

1 eBt

N


1 1 NN eBt
17

(obtained by dividing numerator and denominator by s in 4 and 5)

But we arent finished yet! We had to assume (page 12) that N is never equal to B/s,
ie to N. So we have to think about this.
First, lets ask what happens if N = N ie
N (t) is initially N.
Intuitively, since N is the sustainable population, you would expect that N (t) = constant =
N = N should be possible ! Indeed, substi2
=
BN
sN
and the left
tute N = N into dN
dt
2
side is zero while the right side is BN sN
=

N[B sN] = N[B B] = 0. So we have

N (t) = N (N = N)
18

(9)

Clearly (7) (8) (9) cover all possible values of

N .

Summary: A simple way to improve on Malthus

is to replace his assumption D = constant by
the Logistic Assumption D = sN , s =
constant. If N < B/s = N, then the graph
of N (t) is the S-shape on page 14.
3.3. HARVESTING
A major application of modelling is in dealing
with populations of animals e.g. fish. We want
to know how many we can eat without wiping
them out. Lets build on our logistic model, ie
assume that the fish population would fol19

low that model if we didnt catch any. Next,

assume that we catch E (constant) fish per
year. Then we have:
Basic
dN
= (B sN )N E Harvesting
dt
Model
Well now try to guess what the solutions
should look like. We are particularly interested
in the long term - will the harvesting eventually
exterminate the fish? Consider the function:
F (N ) = (B sN )N E
. The graph of F (N ) can take one of 3 forms.
In the first case, the quadratic
sN 2 + BN E
20

has no solutions, ie
B2
B 4sE < 0 or E >
4s
2

.
Since dN
dt = F (N ), we see that in this case the
21

population always declines. Note that there is

no t-axis in this picture, but you should imagine time passing by thinking of a moving spot
on the graph. Notice that as we move through
t = 1, 2, 3 we have to move to the left since
dN
dt

< 0 always. But | dN

dt | is decreasing, so

the rate of decline slows down as time goes on,

until we pass N =

B
2s .

After that, | dN
dt | increases

and the value of N decreases rapidly to zero.

Congratulations - you have wiped out your fish!
[In drawing this picture, I assumed that N , the
B
, the maxinitial value of N , was greater than 2s

imum point on the graph of F (N ). Note that

22

dN
dt

= F (N ) implies
d2N
dN
0
0
=
F
(N
)
=
F
(N )F (N )
dt2
dt

so watch for points of inflection on the graph

of N (t) at values of N where F 0(N ) or F (N )
vanish.]
Clearly it is not a good idea to harvest at a
rate E >

B2
4s

(Check units). So lets assume that

our fishermen ease off and harvest at a rate E <

B2
4s .

(The special case E exactly equals

23

B2
4s

is clearly impossible in reality, but we will come

back to it later anyway!)
Now the graph of F (N ) is as shown in the
next diagram.

Again, remembering that

that

24

dN
dt

= F (N ), we see

dN
< 0 if 0 < N < 1
dt
dN
> 0 if 1 < N < 1
dt
dN
< 0 if N > 2
dt

and of course dN
dt = 0 at N = 1 and 2 , where

2
1 B B 4Es B B 2 4Es
=
=
2
2s
2s
Now suppose N = N (0) is large, so we start
at point C on the diagram. Then

dN
dt

< 0 so the

fish stocks decline toward 2. Suppose on the

other hand that N is small, but still more than
1. Then we might start at B and the number
of fish will increase until we reach 2. If N
25

is very small, however, then we are at a point

like A, and the fish population will collapse to
zero. So we get a picture like this:

Of course, if N = 1 or 2, then since F (1) =

0 and F (2) = 0,

dN
dt

= F (N ) has solutions

N (t) = 1 and N (t) = 2, the constant solutions. We call 1 and 2 the equilibrium
populations: given a fixed harvesting rate,
26

if the initial population is either 1 or 2, then

in theory the population remains steady which is good! But there is a vast difference between 1 and 2!! Look at the diagram and suppose you have exactly 2 fish.
Now suppose a small number of new fish arrive from somewhere else. Then the diagram
shows that the population will decline back to
2. If some fish go away, the population will
increase back to 2. Of course, such things
happen all the time, so its very good to have
this kind of behaviour! We say that 2 is a
stable equilibrium population. But
now look at 1! If a few more fish arrive,
27

also fine - in fact the population increases

(to 2) but suppose a few fish decide to move
on. Then your fish stocks become extinct!! Note that this can happen though E
is relatively small (we are assuming E <

B2
4s .

We say that 1 is an unstable equilibrium

population. The time required to reach N =
0 is called the extinction time. It can be
computed: since

dN
dt

= N (B sN ) E, we

have:
Z

dt = T =
0

dN
N (B sN ) E

Of course it is very important to know T it is the amount of time you have to save the
situation!
28

Finally, lets consider the third possible graph

for F (N ):

Clearly N =

B
2s

is a solution; it is the only

equilibrium population. If N >

B
2s

(for exam-

ple, at the point ) then the population declines

29

to

B
2s ,

asymptotically. But if we start at the

point , then (since

dN
dt

< 0 everywhere below

the axis - always remember that

dN
dt

= F (N ))

the population will collapse to zero. So we have

an unstable equilibrium at N =

B
2s .

[We only

call it stable if it is stable to perturbations in

both directions!]. The graph of N is as shown.

Clearly, the first and third cases are bad! The

first case was E >

B2
4s .

So we want the second

30

case, with E <

B2
4s ,

and we want stable equi-

librium, that is, a population which fluctuates

around 2 =

B+ B 2 4Es
.
2s

3.4. STEADY GROWTH

There are lots of other assumptions you can
make about B and D. For example, suppose
you are the government of a human population
and you dont like the Malthus model: you want
the population to grow, but NOT exponentially!
Then you can consider a policy like this: you
look at the rate of population growth each year,
and, if you think it is too high, you bring in a
tax on babies in order to discourage people from
having too many. So instead of a constant birth
31

rate per capita of B0, you modify it to

dN
B = B0
,
dt
where is a positive constant. You dont try
to change the death rate per capita D because
that would be political suicide. So now we have,
instead of Malthus,

dN 
dN
= B0
N DN.
dt
dt
This is a separable ODE and you can solve it.
But you can see what is going to happen by
writing the equation as
 dN
1
+
= B0 D.
N
dt
Clearly, when N becomes big, dN/dt becomes
32

approximately constant,
dN
B0 D

.
dt

So with this policy the population will grow, but

only at a constant rate, not at an exponential
rate as the Malthus model would have it. (Here
I am assuming that B0 > D, otherwise you
will find that N decreases rapidly just as in the
Malthus model.)

33

Figure 1: Steady Growth, = B0 D = 1 = N (0).